Computer Science 248/448
- No late homework
- To get a class account, see Marty Gunther
- Turn in assignments using /u/cs248/TURN_IN from class account
- Programming assignments must be in python, and must run on class account
- Homework 1, Due Th 9/7 11am
Implement the HMM decoder using train and test data given in /u/cs248/data/pos.
What is your accuracy on the test file?
- Homework 2, Due Fri 9/15 5pm
Implement the perceptron training algorithm using HMM-based features
and the data set from HW 1.
How does the accuracy on test data accuracy compare to HW 1?
- Homework 3, Due Fri 9/29 5pm
Implement a Kneser Ney trigram language model
using the data set from HW 1.
- Homework 4, Due Th 10/26 in class
Review problems handed out in class
- Homework 5, Due Tu 11/14 11am
Implement IBM Model 1, and plot log probability of training
and test files in /u/cs448/data/mt.
Here is a simple test case.
- Homework 6, Due Tu 11/28 11am
Implement a feed-forward POS tagger using keras.
Network input should be a five word
window centered on the current word.
A good default is that input words are represented with a 100 dimensional embedding,
and the network has one hidden layer with 100 nodes.
You should experiment with different network
To use keras on your class account, set
gildea @ cs rochester edu
November 16, 2017